Implications of spatial summation models for processes of contour perception: a geometric perspective

Abstract

Current models of how the visual system extracts contours from an image are based on its ability to encode relative position, element interactions and some form of spatial summation. The interactive process is usually interpreted as a weighting function inversely proportional to interelement distances. In this paper two particular weighting functions were experimentally compared—an exponential decay process (EDP) and the sinc function. Results indicated that the EDP decay parameter predicted local contour extraction better than the sinc function. The problem of contour extraction is phrased in terms of how the visual system extracts tangent vectors, curvature vectors and invariant vector fields from the stimulus. However it is contended that the underlying processes for these geometric operators reside in the cortical network dynamics that include summation. Results support this process and some considerations are given to the neurophysiological bases for these general functions which not only result in contour extraction but also produce orientation tuning curves and illusions.